Yes, I agree that pcbs won’t be a good idea.
I actually came to the same conclusion - we need a parallelogram as load cell to not mess up the angle. XY will still be affected, but it will be negliglible.
Maybe we need an software option for z-compensation of the extrusion force in z @garethky ? With additional calibration routine? Could be like measure z-offset without extrusion force and with predefined extrusion force to get the factor between extrusion force and z-offset?
Another Idea: One load cell for the hotend to measure z-offset (mechanical lower limit, hotend can only move up) and one load cell for the extruder to measure the pressure advance:
This way we won’t have to deal with the compensation of extruder force.